An address, a few paragraphs, or a printed notice in a photograph can become text you can edit. OCR reads the shapes of letters and produces characters you can copy. It can help with searchable notes or transferring printed material into your own document, provided you compare the result with the source.
The image-to-text tool processes photos and screenshots in your browser. Choose a language, review the recognised text, and download a TXT file. This guide covers preparing the image and checking characters, numbers, and reading order before you rely on that output.
Decide whether you need recognition or a file conversion
A picture of letters contains pixels. OCR attempts to turn them into text characters; an ordinary image-to-PDF conversion does not. If you need to copy a sentence, use recognition. If you only need to send photographed pages in one attachment, JPG to PDF may be enough without extracting any text.
Recognition is not translation. Choosing English does not turn a Bosnian sentence into English. It tells the reader which language to expect. You can edit the recognised text in another application afterwards. A plain-text file does not recreate the source page's colours, fonts, or layout.
Check whether recognition is needed at all. If the text can already be selected and copied from the original document, use that source. Photographing a screen or taking a screenshot of selectable PDF text adds another opportunity for mistakes. For an existing PDF, consider extracting text from PDF first.
Prepare an image with complete, readable characters
Open the picture and inspect the smallest writing. If letters merge because the photo is blurred, try photographing the document again. Increasing the pixel dimensions afterwards does not restore strokes that were never captured. A sharp source is more useful than a larger copy of the same blurry photograph, especially for codes and similar-looking characters.
Keep the camera reasonably square to the paper and light the page without a shadow covering the lines. For an angled photograph, the document scanner can help prepare a flatter view. If the entire page is sideways, use rotate image. Review the corrected copy before feeding it to OCR.
Remove surrounding furniture or application borders with crop image, leaving a little space around the writing. Avoid clipping accents above letters or strokes below the line. The image editing guide explains cropping and resizing when you need to prepare that intermediate copy.
Choose the language printed on the page
Use the combined Bosnian and English option for a document containing both languages. For English-only content, you can choose English. The Bosnian option is also labelled for Croatian and Serbian in Latin script. This setting concerns the text in the image, independently of the language used for the site's buttons.
Cyrillic is not offered as a separate option here. Do not assume that a Serbian Latin setting covers a Cyrillic page. The same caution applies to other languages absent from the menu. If the result contains unrelated characters, check the selected language against the source before repeatedly processing the same image.
Handwriting needs more manual review. A signature, joined letters, or hurried notes are different from evenly printed text. Try a small section and decide whether recognition is helping. A few words you can read yourself may take less time to transcribe than a long sequence of incorrect guesses takes to repair.
Run a small sample before processing more pages
Add an image, review the language setting, and press the recognition button. The first run downloads the program and language data. Keep the tab open during processing. Image contents are processed locally, and completion time also depends on the device doing the work.
For several photographed pages, arrange the file list before starting. Use names that you can match to the source pages. With multiple images, the text result includes filenames separating their recognised content. Remove those labels when moving the text into your document only if you no longer need them for checking.
Begin with a representative page, particularly for a long or poorly photographed document. A successful title page with large lettering does not prove that small notes on later pages will be accurate. Test the hardest section too. That reveals whether you should improve the images before processing the next group.
Keep a clear distinction between an empty result and unreadable output. An empty result means there may be no usable recognised text. A page full of words can still be wrong. Neither condition is fixed merely by changing the filename, and repeated processing of an unchanged source is not a substitute for inspecting it.
Check meaningful characters instead of trusting plausible words
Compare the output with the photograph line by line. For Bosnian content, inspect č and ć, đ and d, and š and s. Names can look plausible with one incorrect letter. The software cannot confirm a person's actual name; the source and your review remain the reference.
Check numbers separately from sentences. Zero can resemble O, and the digit one can resemble lowercase l or uppercase I. Inspect decimal commas, decimal points, and minus signs. A single incorrect character in an address, date, or reference number can change the information even when the surrounding paragraph looks correct.
Read words that your writing application does not underline. An OCR mistake may produce a correctly spelled word with a different meaning. Review line endings in longer passages, where a word can be split or gain an extra space. Do not join every line automatically if the source really begins a new paragraph.
Tables and two-column pages require another check of reading order. Recognised fragments may arrive in a sequence different from the one you expect. If columns become mixed, process them as separate images. For tables, preserve the relationship between each label and its value before rebuilding the cells in your destination document.
Do a final comparison after editing. Corrections can introduce their own mistakes, particularly when several similar numbers occur close together. Keep the image beside the text and check one block at a time. Mark any genuinely unreadable passage in your working copy instead of filling it with a confident guess.
Save the form of text you actually need
TXT suits content you want to paste into notes, a message, or your own template. Download and open it to check that accented characters display correctly. For DOCX, use JPG to Word, which recognises text and saves it in a document. That does not restore the source's original fonts and layout.
For a PDF source, the PDF to Word guide explains the difference between existing text and scanned pages. Avoid extra conversions simply because a familiar extension appears in the filename. Choose the route according to the content inside and the result you need.
Keep the photograph beside the checked text until editing is complete. You can then return to the source if a word or number becomes uncertain. The privacy page explains local processing and other site services. Copying the result into another application is a separate action, so inspect that final version before sharing it.
Apply the steps to your file.
Image to text (OCR)